经络可视化系统中的双目立体视觉定位

Binocular Stereo Vision Localization in Visualization System of Human Meridian

目的针对磁定位经络可视化系统的精度较低并且极易受周围环境影响问题,提出精度及可靠性更高的双目立体视觉定位方法。方法首先采用张正友标定法完成摄像机标定,得到左右摄像机内外参数;其次完成立体匹配,使用双目立体视觉定位系统提取皮肤阻抗检测仪电极板上十字型角点X的亚像素级坐标,并根据对称性和灰度跳变性筛选出X角点,根据唯一性和旋转不变性匹配左右摄像机图像上的X角点;最后利用得到的摄像机内外参数以及X角点匹配的结果计算出经络点的三维坐标并将其投影到场景二维图像上,利用曲线拟合方法连接经络点,从而实现经络的可视化。结果根据标定及匹配结果计算出3个X角点的三维坐标,并结合多通道皮肤阻抗检测仪得到的电极通道,从而得到经络点的三维坐标,最后利用图像融合实现经络的可视化,实验结果精度较高。结论下一步的工作就是优化二值化方法,进一步提高图像融合配准的精度。

Objective As the human meridian visualization system based on a magnetic orientation tracker has lower precision and is easily influenced by the surrounding environment; this paper puts forward a binocular stereo vision location method to improve the precision and reliability. Methods First of all, we used Zhang Zhengyou calibration method to calibrate two cameras to get the inside and outside parameters. Secondly, we matched target comers on the left and right camera image, using these two cameras to extract sub-pixel coordinates of cross comers X which is on the electrode plates of multi-channel meridian impedance detector, filter comer X according to the nature of symmetry and grayscale transitions and match comer X according to the nature of uniqueness and rotation invariance. Finally, used the inside and outside parameters of cameras and the result of match we can calculate the 3-Dimensional （3D） coordinates of the meridian points and project it to the scene on the two-dimensional image, connect meridian points with curve fitting to achieve the visualization of meridian. Results Calculate three 3-Dimensional （3D） coordinates according to calibration and results of match, and get electrode channel with multi-channel meridian impedance detector so as to get 3-Dimensional （3D） coordinates of meridian points. And use image fusion can realize human meridian visualization, and get higher precision. Conclusion Optimizing local extreme value based binarization, and improving precision of image fusion registration would be carried out in the future.